python
def compare_image(expect_path, actual_path, output_path, color_diff_threshold=10.0,max_diff_pixels=100):
# 读取图片
img1 = cv2.imread(expect_path)
img2 = cv2.imread(actual_path)
if img1.shape != img2.shape:
img2 = cv2.resize(img2, (img1.shape[1], img1.shape))
# ---------------------- 颜色差异对比（LAB空间） ----------------------
# 转换为LAB颜色空间（更贴近人类视觉）
img1_lab = cv2.cvtColor(img1, cv2.COLOR_BGR2LAB)
img2_lab = cv2.cvtColor(img2, cv2.COLOR_BGR2LAB)
# 计算色差（欧氏距离）
color_diff = np.sqrt(np.sum((img1_lab - img2_lab) ** 2, axis=2))
mean_color_diff = np.mean(color_diff)
if mean_color_diff > color_diff_threshold:
# 增强色差可视化（将差异值映射到0-255）
color_diff_vis = (color_diff * 255 / np.max(color_diff)).astype(np.uint8)
cv2.imwrite(output_path, color_diff_vis)
raise ValueError(f"颜色差异过大：{mean_color_diff:.2f} > {color_diff_threshold}")
# ---------------------- 像素差异对比（灰度空间） ----------------------
# 灰度化 + 模糊处理
gray1 = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY)
gray2 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
blur1 = cv2.GaussianBlur(gray1, (5, 5), 0)
blur2 = cv2.GaussianBlur(gray2, (5, 5), 0)
# 计算像素差异
diff = cv2.absdiff(blur1, blur2)
_, thresh = cv2.threshold(diff, 25, 255, cv2.THRESH_BINARY)
# 统计显著差异像素数
diff_pixels = np.sum(thresh == 255)
if diff_pixels > max_diff_pixels:
# 标记差异区域（红框）
contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
result = img2.copy()
for cnt in contours:
# if cv2.contourArea(cnt) > min_contour_area:
x, y, w, h = cv2.boundingRect(cnt)
cv2.rectangle(result, (x, y), (x + w, y + h), (0, 0, 255), 2)
cv2.imwrite(output_path, result)
raise ValueError(f"差异像素数超过阈值：{diff_pixels} > {max_diff_pixels}")1、准备好基线图（expect_path）
2、playwright打开页面，截取元素图片（actual_path）
3、直接将基线图和实际元素图片传入对比方法对比，当差异超过预期，则会在实际元素图片上红框标记出差异部分保存到output_path
需要注意的是：
如果是表单截图，则需要等待所有的字段都加载完
如果是图片预览截图，则需要等待图片彻底打开到最大